The short answer is yes, computers can read CAPTCHAs.
But, as mentioned on the Wikipedia article, in most papers/articles discussing this issue, and in the comments above, it depends very much on the CAPTCHA and the algorithm used - some are easy to decode, and other are very hard ( no solutions exist so far ). Font detection is useful for simple CAPTCHAs, but characters can be deformed ( or the image altered ) in ways that render font irrelevant.
reCAPTCHA is discussed in this stackoverflow question, the accepted answer providing a link to site that claims to decode reCAPTCHAs - but it is offline now. The second answer is more informative, mentioning the manipulation of the third annual TIME 100 poll by the infamous Anonymous. Here's a full description of the operation; the salient points being that The Times employed reCAPTCHA after realizing a subversion attempt is under way, which forced Anonymous to vote in an almost brute-force manner:
Update – Just to be perfectly clear,
anon didn’t hack reCAPTCHA. It did
exactly what it was supposed to do. It
shut down the auto voters instantly
and effectively. The only option left
after Time added reCAPTCHA to the poll
was a brute force attack. Ben
Maurer, (chief engineer on reCAPTCHA)
comments on the hack: “reCAPTCHA put
up a hard to break barrier that forced
the attackers to spend hundreds of
hours to obtain a relatively small
number of votes. reCAPTCHA prevented
numerous would-be attackers from
engaging in an attack. In any
high-profile system, it’s important to
implement reCAPTCHA as part of a
larger defense-in-depth strategy”.
As Dr. von Ahn points out “had Time
used reCAPTCHA from the beginning,
this would have never happened — anon
submitted tens of millions of votes
before Time added reCAPTCHA, but they
were only able to submit ~200k
afterwards. And to do this, they had
to resort to typing the CAPTCHAs by
hand!” One thing that Time inc. did
that made it much easier for the
anonymous hack was to allow leave the
door open for cross-site request
forgeries which allowed anon to create
a streamlined poll that never had to
fetch data from Time.com.
They did, however, manage to significantly optimize the process, but that is outside the scope of this question.
Another interesting aspect is mentioned in another answer:
Adaptive Security
reCAPTCHA is a Web service. That means
that all the images are generated and
graded by our servers. In addition to
the convenience that this provides
(you don't have to run costly image
generation scripts on your own
servers), this also provides an extra
level of protection: our CAPTCHAs can
be automatically updated whenever a
security vulnerability is found. For
example, if somebody writes a program
that can read our distorted images, we
can add more distortions in very
little time, and without Web masters
having to change anything on their
side. This is significantly more
secure (and convenient) than having to
re-install a CAPTCHA every time a
vulnerability is found.
This is probably true of most CAPTCHA web services - it is much easier to change the CAPTCHA generating algorithms, then it is for hackers to crack them - other methods would yield better results, such as crowdsourcing to China, or trying to hack the audio CAPTCHA, where available. But a good CAPTCHA has to be readable by humans - which limits the amount of obfuscation that can be employed; a researcher seems to have been able to break reCAPTCHA in 2010, with a success rate of 30%; here's an article, and a video; however, reCAPTCHA seems to have fixed the implementation since.
reCAPTCHA has also allegedly been cracked in February 2011, but no evidence was provided to back the claim.
It should be noted, however, that some popular web sites still use 'broken' CAPTCHAS; for example, PayPals CAPTCHAs can be read by PWNtcha, a program that hasn't been updated for three years.
handy... oh, these crazy german pseudo anglicisms ;)